Actually, this new poll by UMW shows how much difference it makes if you look at registered voters (RVs) vs. likely voters (LVs, however they’re defined by the pollster), and also if you look at a two-way race or throw all the candidates in there. For instance, in this poll, we get:
- Clinton 46%-Trump 38% (two-way, RVs)
- Clinton 46%-Trump 41% (two-way LVs)
- Clinton 38%-Trump 33%-Johnson 10%-McMullin 3%-Stein 2% (five-way, RVs)
- Clinton 40%-Trump 37%-Johnson 8%-McMullin 3%-Stein 1% (five-way, LVs)
Generally speaking, I tend to go with RVs over LVs, especially after Gallup’s 2012 fiasco, because it’s just super-hard to figure out who’s truly “likely” to vote. I mean, in this election, even if Trump supporters right now are saying they’re more enthusiastic, will that last through November 8? Will Clinton’s superior “ground game” get her voters out to the polls, regardless of how “enthusiastic” they are? We don’t know, nor do the pollsters. So, again, I’m skeptical about Likely Voter models.
As for the results from this poll, I’d make a few points. First, we don’t have any previous polling by UMW to compare this one to, so no idea what the trend might be. Also, 538 doesn’t appear to have a grade for UMW, so no idea what the quality or “house effect” of this poll is. Finally, this poll was conducted over six days (September 6-12), a long period of time during which PPP also polled Virginia and came up with a 6-point lead for Clinton in a 5-way race among LVs (not sure what the numbers are for RVs in that poll). So…the UMW poll shows a bit tighter race than PPP (a “B+” pollster in 538) does, but again, I’m not sure what to make of UMW’s quality, house effect or differing top-line number between RVs, LVs, 2-way and 5-way races.
With that, here are a few other interesting findings from the UMW poll:
- President Obama has a 46%-49% (-3 points) approval rating in Virginia among LVs, which is WAYYYY lower than what other polls have shown, making me think this is a more Republican sample than other polls or that Republicans really are far more likely to vote than Dems in Virginia.
- Neither Trump nor Clinton are looked upon favorably in Virginia, with 60% unfavorable for each among LVs.
- Neither candidate is seen as trustworthy either, although believe it or not, pathological liar Trump actually scores slightly higher than Clinton (who scores well on PolitiFact’s scale, while Trump most certainly does NOT score well) on this measure among LVs. Thanks media!
- Clinton scores far higher than Trump when it comes to having “the right kind of temperament and personality to be a good president” (55% of RVs say Clinton does, only 31% of RVs say Trump does); having “strong qualities of leadership” (57% of RVs say Clinton does, only 49% of RVs say Trump does); “understand[ing] the problems of people like you” (47% of RVs say Clinton does vs. just 35% of RVs who think Trump does); being “prepared for the job of president” (57% of RVs say Clinton is vs. only 34% of RVs who say Trump is).
- Most Virginia RVs say that the economy has gotten worse or stayed the same over the past 12 months, which is totally false. Only 34% give the correct answer, which is that the U.S. economy continues to improve. Again, thanks media!
- On immigration, Virginia RVs overwhelmingly (76%-21%) support “providing a way for illegal immigrants currently in the country to gain legal citizenship if they pass background checks, pay fines and have jobs.”
- Tim Kaine gets a 45%-33% approval rating from RVs, which is lower than I’ve seen in other polls.
- By a 50%-40% margin, Virginia RVs say they went for Obama in 2012 (note: Obama won Virginia by 4 points (51%-47%) in 2012, so this is strange. In contrast, Virginia LVs say then went for Obama in 2012 by 3 points (47%-44%), which is a bit lower than the actual results but closer to what actually happened. So maybe the LV model isn’t so bad after all?